def test_base_derivative_non_linear(self):
self.maxDiff = None
series1 = Series.from_array([
[0, 1],
[1, 2],
[2, 4],
[3, 0],
[4, 16],
])
factory = NodeFactory.transformer('test', 'Derivative')
factory.set_param_value('lag', '1')
factory.set_param_value('metric', 'sub')
diff = factory.build()
result, debug_info = diff.transform([series1], False)
expected_series = Series.from_array([
[1, 1],
[2, 2],
[3, -4],
[4, 16],
])
self.assertEqual(list(expected_series.pdseries.index),
list(result.index))
self.assertEqual(list(expected_series.pdseries.values),
list(result.values))
def test_integration_performance(self):
self.maxDiff = None
pipeline = Pipeline.from_json(self.load_json('pipeline'))
input1 = Series.from_array(self.load_json('input1'), unit='ms')
input2 = Series.from_array(self.load_json('input2'), unit='ms')
analyzer = Analyzer(pipeline, True)
result = analyzer.analyze({'1': input1, '2': input2})
# pr = cProfile.Profile()
# pr.enable()
actual_output = result.output_format()
# pr.disable()
# pr.dump_stats('output.prof')
# s = io.StringIO()
# sortby = 'cumulative'
# ps = pstats.Stats(pr, stream=s).sort_stats(sortby)
# ps.print_stats()
# print(s.getvalue())
expected_file = os.path.join(os.path.dirname(__file__),
'resources/perf/expected_output.json')
# Uncomment to fix test
# print(json.dumps(actual_output, indent=2), file=open(expected_file, 'w'))
expected_output = json.loads(Path(expected_file).read_text())
self.assertEqual(expected_output, actual_output)
def test_divide_discard_fill_zero_div(self):
series1 = Series.from_array([[1, 1], [2, 2], [3, 3]])
series2 = Series.from_array([[1, 2], [2, 0], [3, 5]])
factory = NodeFactory.transformer('test', 'Divide')
factory.set_param_value('zero_div', 0)
divide = factory.build()
result, debug_info = divide.transform([series1, series2], False)
expected_series = pd.Series([1 / 2, 0.0, 3 / 5],
index=series1.pdseries.index)
self.assertEqual(list(expected_series.index), list(result.index))
self.assertEqual(list(expected_series.values), list(result.values))
def test_multi_rolling_aggregate_correlation(self):
factory = NodeFactory.transformer('test', 'MultiRollingAggregate')
factory.set_param_value('window', '2')
factory.set_param_value('center', False)
factory.set_param_value('min_periods', '0')
factory.set_param_value('agg_method', 'correlation_pearson')
factory.add_source(InputRef('lhs'))
factory.add_source(InputRef('rhs'))
ram = factory.build()
s1 = Series.from_array([[0, 1], [1, 2], [2, 3], [3, 4], [4, 5]])
s2 = Series.from_array([[0, 1], [1, 2], [2, 3], [3, 2], [4, 1]])
self.case(ram, s1, s2, list(s1.pdseries.index)[1:], [1, 1, -1, -1])
def test_multi_rolling_aggregate_mismatching_start(self):
factory = NodeFactory.transformer('test', 'MultiRollingAggregate')
factory.set_param_value('window', '1')
factory.set_param_value('center', False)
factory.set_param_value('min_periods', '0')
factory.set_param_value('agg_method', 'proportion')
factory.add_source(InputRef('lhs'))
factory.add_source(InputRef('rhs'))
ram = factory.build()
s1 = Series.from_array([[0, 1], [1, 2], [2, 3], [3, 4]])
s2 = Series.from_array([[1, 5], [2, 5], [3, 5]])
self.case(ram, s1, s2, list(s2.pdseries.index), [2 / 5, 3 / 5, 4 / 5])
def test_multi_rolling_aggregate_proportion_window_min_periods_default(
self):
factory = NodeFactory.transformer('test', 'MultiRollingAggregate')
factory.set_param_value('window', '2')
factory.set_param_value('center', False)
factory.set_param_value('min_periods', '')
factory.set_param_value('agg_method', 'proportion')
factory.add_source(InputRef('lhs'))
factory.add_source(InputRef('rhs'))
ram = factory.build()
s1 = Series.from_array([[0, 1], [1, 2], [2, 3], [3, 4]])
s2 = Series.from_array([[0, 5], [1, 5], [2, 5], [3, 5]])
self.case(ram, s1, s2, list(s1.pdseries.index[1:]), [0.3, 0.5, 0.7])
def test_analysis_histogram_heatmap(self):
self.maxDiff = None
series = Series.from_array([
[1628294400, 0],
[1628337600, 0],
[1628380800, 1],
[1628424000, 1],
[1628467200, 1],
[1628510400, 0],
[1628553600, 0],
[1628596800, 0],
])
factory = NodeFactory.detector('test_node', 'SimpleThreshold')
factory.set_param_value('inside', False)
factory.set_param_value('strict', False)
factory.set_param_value('lower', None)
factory.set_param_value('upper', 1)
factory.add_source(InputRef('input'))
node = factory.build()
pipeline = Pipeline([node])
analyzer = Analyzer(pipeline=pipeline, debug=True)
analysis = analyzer.analyze({'input': series})
actual_output = analysis.output_format()
expected_file = os.path.join(os.path.dirname(__file__), 'resources/analysis/expected_histogram_heatmap.json')
# Uncomment to fix test
# print(json.dumps(actual_output, indent=2), file=open(expected_file, 'w'))
expected_output = json.loads(Path(expected_file).read_text())
self.assertEqual(expected_output, actual_output)
def test_exog_seesaw(self):
exog = Series.from_array([
[0, 1],
[1, -1],
[2, 1],
[3, -1],
[4, 1],
[5, -1],
[6, 1],
[7, -1],
[8, 1],
[9, -1],
])
endog = TestSeriesBuilder.linear(10, 0, 1).build()
factory = NodeFactory.transformer('test', 'SARIMAX')
factory.set_param_value('p', '1')
factory.set_param_value('d', 1)
factory.set_param_value('output', 'resid')
sarimax = factory.build()
result, debug_info = sarimax.transform([endog, exog], True)
expected_series = [0] * 8
actual_series = list(result.values)
for i in range(0, len(expected_series)):
self.assertAlmostEqual(expected_series[i], actual_series[i], 2)
# With debug info
self.assertEqual(set(['summary', 'offset_start', 'exog_coeff_x1']),
set(debug_info.keys()))
def __init__(self, methodName):
unittest.TestCase.__init__(self, methodName)
self.series = Series.from_array([
[0, 0],
[1, 1],
[2, 1],
[3, 0],
])
def test_anomalies_from_mock_edge_case(self):
test_series = Series.from_array([[0, 1], [1, 0], [2, 1]])
anomalies = TestCase.anomalies_from_mock_series(test_series)
self.assertEqual(2, len(anomalies))
self.assertEqual(test_series.pdseries.index[0], anomalies[0].start)
self.assertEqual(test_series.pdseries.index[0], anomalies[0].end)
self.assertEqual(test_series.pdseries.index[2], anomalies[1].start)
self.assertEqual(test_series.pdseries.index[2], anomalies[1].end)
def case(self, node, fst_anomalies, snd_anomalies, expected_output,
expected_debug):
fst_node = Node('fst')
snd_node = Node('snd')
[a.set_source_node(fst_node) for a in fst_anomalies]
[a.set_source_node(snd_node) for a in snd_anomalies]
fst_series = Series.from_array([[0, 0], [1, 0], [2, 0], [3, 0], [4, 0],
[5, 0], [6, 0]])
snd_series = Series.from_array([[0, 0], [1, 0], [2, 0], [3, 0], [4, 0],
[5, 0], [6, 0]])
fst_input = NodeResult(None,
None,
output_series=fst_series,
anomalies=fst_anomalies)
snd_input = NodeResult(None,
None,
output_series=snd_series,
anomalies=snd_anomalies)
result = node.execute([fst_input, snd_input], True)
expected_display_series = {
'input_1': fst_series,
'input_2': snd_series
}
self.assertEqual(expected_display_series, result.display_series())
actual_anomalies = list(
map(lambda a: a.output_format(), result.anomalies))
self.assertEqual(expected_output, actual_anomalies)
self.assertEqual(expected_debug, result.debug_info)
# Check no anomaly sources outside of original anomalies
all_anomaly_sources = set()
for anomaly in result.anomalies:
for source_anomaly in anomaly.source_anomalies:
all_anomaly_sources.add(source_anomaly)
self.assertEqual(
set(), all_anomaly_sources - set(fst_anomalies + snd_anomalies))
def build_series(self):
series = Series.from_array([
[0, 1],
[1, 1],
[2, 1],
[3, 3],
[4, 3],
[5, 1],
[6, 1],
[7, 1],
])
return series
def case(self, node, fst_anomalies, snd_anomalies):
self.maxDiff = None
fst_node = Node('fst')
snd_node = Node('snd')
[a.set_source_node(fst_node) for a in fst_anomalies]
[a.set_source_node(snd_node) for a in snd_anomalies]
fst_series = Series.from_array([
[0, 0],
[1, 0],
[2, 0],
[3, 0],
[4, 0],
[5, 0],
[6, 0]
])
snd_series = Series.from_array([
[0, 0],
[1, 0],
[2, 0],
[3, 0],
[4, 0],
[5, 0],
[6, 0]
])
fst_input = NodeResult(None, None, output_series=fst_series, anomalies=fst_anomalies)
snd_input = NodeResult(None, None, output_series=snd_series, anomalies=snd_anomalies)
result = node.execute([fst_input, snd_input], False)
expected_display_series = {'input_1': fst_series, 'input_2': snd_series}
self.assertEqual(expected_display_series, result.display_series())
return result.anomalies
def test_seesaw(self):
series = Series.from_array([[0, 1], [1, -1], [2, 1], [3, -1], [4, 1],
[5, -1]])
factory = NodeFactory.transformer('test', 'SARIMAX')
factory.set_param_value('p', '1')
factory.set_param_value('output', 'resid')
sarimax = factory.build()
result, debug_info = sarimax.transform([series], True)
expected_series = [0] * 5
actual_series = list(result.values)
for i in range(0, len(expected_series)):
self.assertAlmostEqual(expected_series[i], actual_series[i], 2)
def test_metrics_case1(self):
test_series = Series.from_array([[0, 0], [1, 1], [2, 0], [3, 0],
[4, 0], [5, 1], [6, 1], [7, 0]])
actual_anomalies = [
Anomaly.from_epoch(1, 1),
Anomaly.from_epoch(2, 2),
Anomaly.from_epoch(4, 4)
]
analyzer = MockAnalyzer(actual_anomalies)
expected_anomalies = TestCase.anomalies_from_mock_series(test_series)
expected_analysis = Analysis(test_series, None, expected_anomalies,
False)
test = TestCase("testcase1", test_series, analyzer, expected_analysis)
result = test.run()
anomaly_metrics = result.anomaly_metrics
expected_tp = mcb.epochs_to_timestamp([1])
expected_fp = mcb.epochs_to_timestamp([2, 4])
expected_tn = mcb.epochs_to_timestamp([0, 3, 7])
expected_fn = mcb.epochs_to_timestamp([5, 6])
self.assertEqual(expected_tp, anomaly_metrics.tp)
self.assertEqual(expected_fp, anomaly_metrics.fp)
self.assertEqual(expected_tn, anomaly_metrics.tn)
self.assertEqual(expected_fn, anomaly_metrics.fn)
self.assertEqual(1, anomaly_metrics.tp_count())
self.assertEqual(2, anomaly_metrics.fp_count())
self.assertEqual(2, anomaly_metrics.fn_count())
self.assertEqual(3, anomaly_metrics.tn_count())
expected_tp_ranges = mcb.epochs_to_ranges([[1, 2]])
expected_fp_ranges = mcb.epochs_to_ranges([[2, 3], [4, 5]])
expected_fn_ranges = mcb.epochs_to_ranges([[5, 7]])
self.assertEqual(expected_tp_ranges, anomaly_metrics.tp_ranges)
self.assertEqual(expected_fp_ranges, anomaly_metrics.fp_ranges)
self.assertEqual(expected_fn_ranges, anomaly_metrics.fn_ranges)
self.assertAlmostEqual(0.33, anomaly_metrics.precision(), 2)
self.assertAlmostEqual(0.33, anomaly_metrics.recall(), 2)
self.assertAlmostEqual(0.33, anomaly_metrics.f1(), 2)
def test_base_derivative_lag2(self):
series1 = TestSeriesBuilder.linear(5, 0, 1).build()
factory = NodeFactory.transformer('test', 'Derivative')
factory.set_param_value('lag', '2')
factory.set_param_value('metric', 'sub')
diff = factory.build()
result, debug_info = diff.transform([series1], False)
expected_series = Series.from_array([
[2, 2],
[3, 2],
[4, 2],
])
self.assertEqual(list(expected_series.pdseries.index),
list(result.index))
self.assertEqual(list(expected_series.pdseries.values),
list(result.values))
def build(self):
return Series.from_array(self.data, self.unit)